Work

Evaluating the Accuracy of Open-Source AI Text Detectors

Year: 2026

Type: preprint

Abstract: This paper evaluates the performance and accuracy of automated Artificial Intelligence (AI) text detection tools. As Large Language Models (LLMs) become common, identifying machine-generated text is a critical area of data evaluation. In this study, a dataset of 10 human-written paragraphs and 10 ChatGPT-generated paragraphs was evaluated using a free online AI detector. The detector achieved an overall classification accuracy of 90%, proving highly effective but showing vulnerabilities to minor text editing.

Source: Zenodo (CERN European Organization for Nuclear Research)

DOI: 10.5281/zenodo.24038119

Author: Das, Debankan

Language: English

Topic: Authorship Attribution and Profiling

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: green

OpenAlex ID: W7225459135

Cites:

Cited by:

Related to: 10